360 Secure SkillHub
360 安全技能中心

Veto Approval (Human-in-the-Loop)

通过导入 Python 模块调用外部审批服务,在删除数据库等高风险动作前向指定邮箱发送带批准/拒绝按钮的邮件,并每 5 秒轮询直到人工给出决定,返回 True 或 False 以决定是否继续执行。

mrchop
mrchop
热度 36.92
最近更新 2026-05-11 17:36:07
分类:安全依赖:代码执行 / 第三方 API应用场景:工作流 / 自动化脚本 / 任务管理

安装技能

复制下面的话术发给你的 Agent,它会通过 secure-skills 检索并安装该技能。

执行命令安装此 skill: npx secure-skills-cli add veto-approval
Metadata
name
Veto Approval
description
Pause your agent for human approval before high-risk actions.
parameters
{"action_title":"string (e.g. \"Delete user database\")","user_email":"string (The approver's email)","context":"string (Optional context for the email)"}

Veto Approval Skill

This skill allows your agent to send a 1-click approval email to you before proceeding with sensitive tasks.

Setup

  1. Get your API Key at vetoapi.com
  2. Add VETO_API_KEY to your environment.

Usage

from veto_skill import ask_human_permission

approved = ask_human_permission(
    action_title="Send invoice for $5,000 to Client X",
    user_email="boss@company.com",
    context="Monthly retainer — Q1 2026"
)

if approved:
    send_invoice()
else:
    print("Action blocked by human.")

How it works

  1. Your agent calls ask_human_permission()
  2. An approval email with Approve and Reject buttons is sent to user_email
  3. The function blocks until a decision is made (polls every 5 seconds)
  4. Returns True if approved, False if rejected